Linux network down or flaky: diagnose and fix it from the terminal

What this covers

When a Linux box loses connectivity or the network turns slow and unreliable, the fix is a systematic walk up the stack: interface, routing, path, then connectivity. This guide uses only verified commands to find where the failure is and confirm the repair.

Symptom

Check your interfaces have an address

Start at the bottom of the stack: a network problem is often just an interface that is DOWN or missing its IP address. A compact table of every interface with its state and addresses tells you in seconds whether the link itself is the problem. If your interface shows DOWN or has no address, the fix is connectivity, not routing.

$ ip -br a

Check which route the destination actually takes

If the interface looks fine, ask the kernel directly which interface, source address and gateway it would use to reach the destination. This reveals wrong default gateways or traffic leaving on the wrong interface — a common cause when some hosts work and others do not. You succeeded when the output names a sane gateway for your destination.

$ ip route get 192.168.1.10

Watch for intermittent loss over time

A single ping tells you almost nothing about a flaky connection. Run a continuous ping with timestamps to catch packet loss and latency spikes as they happen, and to correlate them with events (a reboot, a heavy download, a time of day). If replies stop or loss appears, you have reproduced the symptom and know when it strikes.

$ ping -D 8.8.8.8

Find where along the path packets die

When the destination is reachable but slow or lossy, you need to know whether the problem is in your network, your ISP, or closer to the server. mtr keeps probing every hop and shows live loss and latency per router, so you can point at the exact hop where quality collapses. Loss that appears at one hop and stays through the rest of the path locates the culprit.

$ mtr -rw example.com

Reconnect the link if Wi-Fi is down

If the diagnose steps showed a DOWN interface or a missing connection — common on headless servers over SSH — rebuild the link with NetworkManager. Connecting from the terminal re-establishes the association and saves the profile, so the fix also survives reboots. Success: the interface comes up and the earlier checks now show an address.

$ nmcli device wifi connect "MySSID" password "MyPassword"

$ nmcli connection delete "MySSID"

Confirm the full path is healthy again

After any fix, run a traceroute to the destination: every hop should answer within reasonable round-trip times and the final hop should be the host you care about. Comparing against the mtr output from the diagnose phase shows whether the bad hop is gone or improved. Consistent, complete replies mean the path is verified end to end.

$ traceroute example.com

Verify outbound connectivity from the outside

Finally, confirm you can actually reach the internet, not just your gateway. Printing your outbound public IPv4 address proves DNS, routing and NAT all work end to end. If it returns an address, your connection is fully restored — and if a VPN is involved, it also tells you whether traffic is really leaving through it.

$ curl -s https://api.ipify.org

How to confirm it worked

A clean traceroute to your destination with consistent replies, plus a successful public IP lookup, confirms routing and outbound connectivity are restored.

Undo everything

Every command above can be reverted. Run these in order to undo all of it.

$ nmcli connection delete "MySSID"

Review before running — some steps may not be independent.

Sources

• ip(8) — Linux manual page
  man7.org · technical · 2026-09-17
  Evidence: Documents `ip -br addr` as the brief, one-line-per-interface output format of the ip command
• ip-route(8) — main IPv4/IPv6 routing table management
  Debian manpages · community · 2026-09-17
  Evidence: Page documents the 'get' subcommand ('ip route get' — get a single route) and route attributes like via/dev/src; confirms 'ip route' and 'get' appear in the text.
• ip-route(8) — Ubuntu man page
  Ubuntu manpages · community · 2026-09-17
  Evidence: Man page contains the 'get' subcommand description showing it performs a route lookup and returns the resolved via/dev/src values.
• ping(8) — send ICMP ECHO_REQUEST to network hosts
  man7.org · technical · 2026-09-17
  Evidence: Documents -D to print timestamps before each line and -i interval / -c count semantics.
• mtr(8) — combines the functionality of traceroute and ping
  Ubuntu Manpages · technical · 2026-09-17
  Evidence: Documents mtr as traceroute+ping combined, and the -r report mode with -w wide format
• MTR — My Traceroute
  BitWizard · vendor-doc · 2026-09-17
  Evidence: Official mtr project page describing its real-time route/loss statistics display
• nmcli(1) — command-line tool for controlling NetworkManager
  Ubuntu Manpages · technical · 2026-09-17
  Evidence: Documents `nmcli device wifi connect <SSID> password <pass>` and `nmcli device wifi list` for scanning
• traceroute(8) — print the route packets trace to network host
  Arch Linux Manpages · technical · 2026-09-17
  Evidence: Documents traceroute's default UDP probing, TTL increments, and output format with hop latency

Last verified: 2026-09-21

Related guides